econstor Make Your Publications Visible.

Transcription

1 econstor Make Your Publications Visible. A Service of Wirtschaft Centre zbwleibniz-informationszentrum Economics Rommel, Tobias; Schaudt, Paul Working Paper First Impressions: How Leader Changes Affect Bilateral Aid CESifo Working Paper, No Provided in Cooperation with: Ifo Institute Leibniz Institute for Economic Research at the University of Munich Suggested Citation: Rommel, Tobias; Schaudt, Paul (2016) : First Impressions: How Leader Changes Affect Bilateral Aid, CESifo Working Paper, No This Version is available at: Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence.

2 First Impressions: How Leader Changes Affect Bilateral Aid Tobias Rommel Paul Schaudt CESIFO WORKING PAPER NO CATEGORY 2: PUBLIC CHOICE AUGUST 2016 An electronic version of the paper may be downloaded from the SSRN website: from the RePEc website: from the CESifo website: Twww.CESifo-group.org/wpT ISSN

3 CESifo Working Paper No First Impressions: How Leader Changes Affect Bilateral Aid Abstract We present a new mechanism to explain politically induced changes in bilateral aid. We argue that shifts in the foreign policy alignment between a donor and a recipient country arising from leadership changes induce reallocation of development aid. Utilizing data from the G7 and 133 developing countries between 1975 and 2012, we show that incoming leaders in recipient countries that politically converge towards their donors receive more aid commitments, compared to those that diverge. Additionally taking donor leader change into account, we find that incumbent recipient leaders have an opportunity to get even more aid when political change in donor countries moves them closer to the donor s foreign policy position. Thus, leadership turnover in recipient and donor countries constitutes a potential breaking point in foreign relations. Otherwise inconsequential deviations in voting alignment become highly consequential for aid provision. JEL-Codes: D720, F350, F530, O190. Keywords: dyadic leader change, UNGA voting realignment, development aid. Tobias Rommel University of Zurich Zurich / Switzerland rommel@ipz.uzh.ch Paul Schaudt University of Hannover Hanover / Germany schaudt@glad.uni-hannover.de Version: August 4, 2016 We thank Axel Dreher, Martin Gassebner, Krisztina Kis-Katos, Tommy Krieger, Mariana Lopes da Fonseca, Christoph Mikulaschek, Sibel Oktay, Ari Ray, Stefanie Walter, Wen-Chin Wu, as well as participants at the CESifo Workshop on Political Economy (Dresden, 2015), the Annual Conference on the Political Economy of International Organizations (Salt Lake City, 2016), the Annual Meeting of the European Public Choice Society (Freiburg, 2016), the Annual Conference of the Midwest Political Science Association (Chicago, 2016), the Annual International Conference of the Research Group on Development Economics of the German Economic Association (Heidelberg, 2016), and the Workshop on Political Economy of the Silvaplana Political Economy Group (Pontresina, 2016) for valuable and helpful comments. We also thank Erasmus Kersting, Christopher Kilby, and Shu Yu for making their data available.

4 1 Introduction Official development aid (ODA) is an important source of financial liquidity for developing countries. If funds run dry, these countries face severe economic repercussions. 1 As aid is in part granted on political grounds, both the size and volatility of aid flows are subject to politics. Long-term relations such as colonial ties (e.g., Alesina and Dollar, 2000) and short-term shifts in the political importance of recipients such as membership in the United Nations Security Council (UNSC) (e.g., Kuziemko and Werker, 2006; Dreher, Sturm, and Vreeland, 2009) affect bilateral ODA flows. Apart from a recipient s international standing, its political positions matter as well. Disagreement between donors and recipients on policies significantly lowers aid flows, especially if issues are highly relevant for donors (Andersen, Hansen, and Markussen, 2006; Dippel, 2015; Vreeland and Dreher, 2014). Donors even adjust access to liquidity strategically in order to influence elections in recipients countries. They increase bilateral aid to political friends during election years, thereby bolstering re-election prospects, while they decrease aid to political opponents (Faye and Niehaus, 2012). 2 Given the fact that donors try to keep their friends in power and actively foster political change in hostile countries, it is surprising that we know only little about the impact of leadership turnover on aid allocation: How do donors adjust aid provision following leader change? Our paper provides an answer to this question. Leadership turnover in both recipient as well as donor countries is a source of uncertainty concerning future behavior in the international arena. Since the pursuit of foreign policy is usually the prerogative of the executive branch, leader change opens the door for large-scale policy change. Although the potential for policy change is high, re-alignment can go in both directions. New leadership does not automatically guarantee improved bilateral relations between donors and recipients. Hence, the consequences of leader change for aid allocation are ambiguous ex ante. Because uncertainty rises, we argue that donors take foreign policy positions announced by recipients under increased scrutiny. Thus, shifts in foreign policy following leader change work as an important source of information on which donors base their decisions regarding aid allocation. What is more, we argue that the effect of political re-alignment on aid allocation is not only present in case of leadership change in recipient countries, but is also visible following leadership change in donor countries. Given that political relationships between states are reciprocal, changes in the head of executive of donor countries similarly increase uncertainty by discounting past behavior and therefore expecta- 1 Not only are aid fluctuations harmful for the economic prosperity of least developed countries (Ebeke and Ölçer, 2013), but they also escalate the risk of violent civil conflict (Nielsen et al., 2011). Aid volatility increases the probability of regime survival (Morrison, 2009; Kono and Montinola, 2009), yet access to easy money simultaneously tends to weaken the democratic quality of political institutions (Bueno de Mesquita and Smith, 2010, 2013). 2 Similarly, the United States use their weight in the International Monetary Fund to provide loose conditions on credits (Dreher and Jensen, 2007) and in the World Bank to provide quicker loan disbursement (Kersting and Kilby, 2016) for political friends in the run-up to elections. 1

5 tions about future relations. Hence, new donor leaders base aid disbursement on the foreign policy changes of recipient country governments. Our argument thus predicts that sizable reallocation of development aid occurs after either recipient or donor leader change. Yet, the direction should depend on the foreign policy shifts that recipient countries send towards donors. Leaders who signal political accord receive more aid; countries receive less aid if a leader signals political animosity. To capture foreign policy shifts, we rely on comparable measures of voting alignment in the UNGA (Voeten, 2000). Voting in international organizations is a very cost-effective way for donors to observe political alignment or dis-alignment of their recipients. Accordingly, UNGA voting patterns have frequently been used to proxy for political closeness between countries (e.g., Thacker, 1999; Barro and Lee, 2005; Bailey, Strezhnev, and Voeten, 2015). Indeed, studies suggest that changes in heads of executive make a decisive difference when it comes to foreign policy proximity (Dreher and Jensen, 2013; Mattes, Leeds, and Carroll, 2015). Yet, research has focused exclusively on either leadership changes in recipient countries only or monadic position changes. We extend this research and argue that leader changes in both recipient and donor countries affect political friendship. Building on this insight, we scrutinize the effect that a change in voting alignment has on aid commitments. We focus on alignment changes that occur after leadership change in a dyadic donor-recipient leader pair, between leaders from the G7 and developing countries. Covering 133 recipient countries from 1975 to 2012, our analysis shows that yearly alterations of foreign policy alignment have no significant effect on aid commitments, unless occurring after leadership turnover. In line with our argument, the adjustment of foreign policy objectives after leader changes has a huge impact on aid commitments. Donors reward political convergence and punish divergence. Nevertheless, these effects are different in substantial terms. Our findings suggest that leader changes in donor countries represent a window of opportunity that recipients can use to attract windfall gains in development aid, while recipient leader changes open predominantly a window of dis-opportunity to forgo aid cutbacks. Focusing exclusively on monadic leadership changes in recipient countries is, thus, not able to capture essential variation in the allocation of aid induced by leadership changes. Taken together, recipient country leaders have to decide early on, how to align themselves with their international aid providers. We proceed as follows: Section 2 presents our theoretical argument linking dyadic leadership change, political alignment, and aid allocation. Section 3 describes the data. Section 4 discusses our empirical strategy and results. Section 5 presents robustness tests. Section 6 concludes. 2 Theoretical Argument How does leadership change affect aid allocation? We argue that leadership turnover, both in recipient as well as donor states, opens the door for large-scale policy changes. 2

6 Such changes increase uncertainty about future behavior in the international arena. To mitigate uncertainty, donor countries observe foreign policy shifts following leader change in recipient countries and utilize them as signals before making decisions about aid allocation. Similarly, new donor leaders base their aid allocation on the foreign policy signals recipient countries send ex post. In both cases, they reward alignment and punish dis-alignment resulting in reallocations of development aid. 2.1 Leadership Change and the Foreign Policy Agenda Why does leadership change matter when it comes to foreign policy alignment? Realist accounts of international politics argue that political proximity between countries is primarily determined by the anarchic structure of the international system (Waltz, 1959, 1979). According to this perspective, the interests of nation states are presumed to be rather constant over time (Morgenthau, 1948). Looking at UNGA voting patterns, this notion manifests itself, for example, through the importance of voting blocs (Kim and Russett, 1996). Countries ally themselves with other countries in their region, with countries that they have formed military alliances with (Leeds and Mattes, 2007), and with countries that belong to the same international regime such as the Commonwealth of Independent States (Hansen, 2015). Their foreign policy positions also tend to be closer to countries that resemble themselves in terms of economic development and type of political system. Accordingly, democracies vote more in line with other democracies on soft-power (Voeten, 2000, 2004) as well as on hard-power issues, such as self-defense in the context of threats emanating from terrorist groups (Hillman and Potrafke, 2015). What is more, democracies are easier to bribe, because donor electorates find it more acceptable to provide additional funds to democracies than autocratic leaders (Carter and Stone, 2015). While these arguments can explain general alignment tendencies in the long run, they fail to explain short-term shifts in political relationship between countries. Voeten (2000), for example, finds that voting alliances are much more ad hoc, issuebased and fragile since the end of the Cold War. Hug and Lukács (2014) show that country preferences tend to trump voting blocs in controversial votes in the United Nations Human Rights Council. To explain short-term variation, we stress the role that heads of executive play in pursuing foreign policy goals and argue that leadership changes open up opportunities for structural breaks in foreign policy alignment for two reasons. First, leadership turnover is accompanied by alterations in domestic support groups and, second, it resets personal relationships between leaders. Such instances are thus potential breaking points in bilateral relations. Assuming that leaders want to stay in power, the pursuit of foreign policy objectives is curtailed internally (Moravcsik, 1997; Putnam, 1988). Domestic support groups, powerful interest groups, and public opinion influence the re-election prospects of democratic leaders and the power maintenance of autocratic leaders alike. Therefore, changes in foreign policy objectives are due to changes in the 3

7 domestic support group (Jacobs and Page, 2005; Aldrich et al., 2006). A democratically elected vote- and office-seeking government faces constraints by its electorate and therefore chooses its foreign policy in accordance with the preferences of this part of the populace (Moravcsik, 1993). In autocracies the population is generally not eligible to vote or their vote does not make a decisive difference. Nonetheless, leaders in dictatorships are similarly constrained by core support groups like the regime party, the royal family, or the military (Cheibub, Gandhi, and Vreeland, 2010; Geddes, Wright, and Frantz, 2014) and pursue foreign policy goals that are consistent with these interests (Weeks, 2012; Way and Weeks, 2014; Mattes and Rodriguez, 2014). Hence, interests of influential groups shape leaders foreign policy agenda in both democracies and autocracies. These agenda are then presented in the international arena. In line with this, Dreher and Jensen (2013) find that leader changes increase voting alignment of other countries with the United States on key votes. This implies that a country is more likely to agree with the United States on important issues if new leadership reshapes foreign policy. Mattes, Leeds, and Carroll (2015) add that leader changes lead to a shift of ideal points in UNGA voting only if accompanied by simultaneous changes in the domestic support group. They also show that democracies have more stable ideal points than autocracies. To the contrary, leader effects are stronger in autocratic regimes, because the support base is narrower than in democracies and changes in the domestic support groups can be more pronounced (see also Bueno de Mesquita et al., 2003). Finally, Mattes, Leeds, and Carroll (2015) highlight that changes in the foreign policy agenda of a country are more abrupt between administrations than within. Although leaders have to consider domestic constraints in order to hold on to power, they still have considerable leeway when pursuing foreign policy. Politicians might have different visions or weigh the interests of some groups more than of others. Because governments still have some leverage when it comes to satisfying the foreign policy preferences of their core support groups, personal relationships between statesmen similarly affect the chances for cooperation and confrontation in international relations alike (Hermann, 1980, 1990). Consider as an example the 2015 nuclear agreement with Iran, which was possible after Hassan Rouhani s inauguration and seemed unthinkable under his predecessor Mahmoud Ahmadinejad. Likewise, such an agreement was unthinkable under previous President George W. Bush. The fact that President Barack Obama, in a system with checks and balances like in the United States and with large parts of the US Congress opposing a deal, could arrange an agreement even further highlights the effect of specific leaders. Hence, even when considering all domestic constraints that policymakers face when conducting foreign policy, personal relationships between statesmen still matter. 3 In summary, leader change opens up opportunities for a reorientation of for- 3 In addition, leadership changes affect various other outcomes, such as trade or monetary policy (McGillivray and Smith, 2004) or democratization and conflict (Jones and Olken, 2005, 2009). 4

8 eign policies for two reasons: either changes in the domestic support group call for a new foreign policy agenda, or the individual relationship and interaction between leaders induces new dynamics in bilateral relations between countries. Either way, because the pursuit of foreign policy objectives is the executive branch s prerogative, leadership turnover in each country of a dyadic country pair should serve as a potential breaking point for foreign policy proximity between countries. 2.2 Foreign Policy Shifts and Aid Allocation International relations can be characterized as a set of bilateral relations between any two countries. Accordingly, the very nature of political alignment is reciprocal. Given that donor countries care about which recipient leader is in power and have vested interests in political alignment with developing countries, leadership turnover in recipient countries increases uncertainty about future behavior in international politics as it sets the stage for new foreign policy agendas. Because a new leader in a recipient country has the potential to change bilateral relations, the leader in a donor country faces increasing uncertainty about the behavior of the recipient leader in the aftermath of leadership change; and vice versa. After inauguration, a new recipient country government can adjust its foreign policy towards the donors in three ways: keep relations unchanged, converge towards a common ideal position on international issues, or diverge. As a reaction to changes in political alignment, we argue that a donor country in turn possesses two policy options: reward political friends with external revenues or deprive opponents of political and economic benefits. Given that donors have an incentive to bind promising new leaders early on, they have a rationale to reward them with additional aid. Contrariwise, a donor makes it hard for a new leader by cutting aid if they perceive them as hostile. Whether a country under new leadership is a political friend or foe is difficult to evaluate in advance (Fearon, 1995, 1997). Relying on ex ante characteristics such as the foreign policy stances of leaders in the run-up to elections may provide an incomplete picture of an administration s foreign policy agenda. Past observed behavior should be heavily discounted as governments have private information that shape their foreign policy preferences, as well as incentives to conceal their true intentions. Additionally, audience costs change in conjunction with leadership turnover, effectively changing incentive structures for the leader after an election. Lastly, the new leader may only imperfectly be bound to path-dependency, or even have come to power by opposing the existing policy platform. Hence, the reaction of the donor hinges on the ex post conduct of the new leadership in the recipient country. We argue that donor countries observe the behavior of new recipient country leaders during their first year in office, for example via voting alignment in the UNGA. Such votes cover a wide array of issues that allow political actors to estimate alignment tendencies and are thus a record of how the state wants to be seen by 5

9 others, the international norms it finds acceptable, and the positions it is willing to take publicly (Mattes, Leeds, and Carroll, 2015, 283). Voting in line with (or against) a donor s interests thus constitutes a cost effective source of information that the donor can observe and use to determine if the other leader is more likely to be a friend or foe in the future. Note that signaling is especially important if there is no prior observable behavior of an actor. Summing up, leadership change itself should not necessarily alter the allocation decision of the donor. Rather, the donor s willingness to provide ODA is shaped by the initial foreign policy positions that a new recipient leader takes. H1: The effect of recipient country leadership change on aid flows is conditional on the political alignment direction new leaders establish towards their donor, during their first year in power. Alignment with the donor increases aid flows; dis-alignment decreases aid flows. Nonetheless, the importance of signaling does not only originate from recipient country leadership changes. If a donor country leader enforces a new set of foreign policy objectives, the repercussions influence a recipient country s ability to pursue and implement its own policy goals. In other words, leadership changes in donor countries themselves shape bilateral foreign policy proximity. Thus, the pursuit of foreign policy goals is further confined by external constraints that arise from the behavior and power of other countries. In essence, both leaders matter for bilateral relations between countries. What is more, reacting to changes in donor countries might be in the interest of recipient countries. Internal constraints are fixed in the short run. Leaders are usually not able to change their support group the electorate in democracies or the selectorate in autocracies because the associated costs endanger their hold on power. Changes in the donor s foreign policy that emanate from leadership change thus open a window of opportunity for recipient countries to change bilateral relations, as external constraints on foreign policy decrease. Consider that newly elected US presidents attempt to accomplish international success rather quickly. For example, Barack Obama vowed to reset relations with the Middle East and reduce US interference in his Cairo speech, held shortly after his 2009 inauguration (New York Times, 2009). In these instances, donor leaders consider the reactions from the developing world as approval or dis-approval. A recipient country can either show willingness to work together or take a stance and openly oppose the new foreign policy agenda of the donor. In this sense, a change in donor leadership can provide other countries with the opportunity either to reset relations or withdraw loyalty, respectively. If leaders welcome a new president and signal that they will work with them, they receive additional aid as part of a charm offensive. If a new leader in a donor country receives hostile signals from a recipient country s political leadership, aid flows conversely decrease. In both cases we argue that first impressions matter a great deal and should influence the allocation of aid. 6

10 H2: Recipient country convergence towards a donor s foreign policy position after donor country leadership change increases aid flows; divergence decreases aid flows. In summary, leadership changes in both the recipient and the donor country reset personal relationships and the domestic constraints on leaders and open windows of opportunity to fundamentally change foreign policy. In such situations, uncertainty in the bilateral relations between a donor and a recipient country rises and donor leaders make aid allocation decisions depending on ex post changes in foreign policy positions of recipient countries. Because donor countries have vested interests in political alignment, they reward political alignment and punish dis-alignment. 3 Data and Operationalization Our dependent variable is official development aid. 4 In line with Faye and Niehaus (2012), we use ODA commitments instead of disbursements, since disbursements in a given year might originate from projects granted earlier. Commitments on the other hand are targeted to a specific country in a given year. Hence, we can directly link them to shifts in political alignment between countries arising from leadership turnover. We take ODA commitments from the Development Action Committee (DAC) database of the OECD (2015). Because aid commitments are highly skewed, we use log-transformed values. We focus mainly on country dyads with positive aid flows to avoid arbitrary log-transformations. Nevertheless, we control for the inclusion of zeros as well as for selection effects in the robustness section. 5 The first independent variable is dyadic leadership change. To measure this, we use data from the updated Archigos dataset (Goemans, Gleditsch, and Chiozza, 2009) to identify the heads of executive of each recipient and donor country. We code a change in leadership if the leader of country i in year t differs from the leader of country i in year t 1. If several leaders were in power in a country during a given year, we focus on the leader that has spent the highest fraction of days in office over the course of the respective year. As such, we assume that more days in office 4 We define ODA as those flows to countries and territories on the DAC list of ODA recipients and to multilateral institutions which are: i. provided by official agencies, including state and local governments, or by their executive agencies; and ii. each transaction of which: a) is administered with the promotion of the economic development and welfare of developing countries as its main objective; and b) is concessional in character and conveys a grant element of at least 25 per cent (calculated at a rate of discount of 10 per cent) (OECD, 2015). Over the years the DAC has refined the ODA reporting rules to ensure accuracy and consistency among donors. The boundary of ODA has been carefully delineated, including: 1. Military aid: No military equipment or services are reportable as ODA. Anti-terrorism activities are also excluded. The cost of using donors armed forces to deliver humanitarian aid is eligible. 2. Peacekeeping: Most peacekeeping expenditures are excluded in line with the exclusion of military costs. Some closely defined developmentally relevant activities within peacekeeping operations are included. 3. Nuclear energy: Reportable as ODA, provided it is for civilian purposes. 4. Cultural programs: Eligible as ODA if they increase cultural capacities, but one-off tours by donor country artists or sportsmen, and activities to promote the donors image, are excluded. 5 Note that 23% of the bilateral aid flows are zero. This is mainly driven by the complete absence of development cooperation between Japan and several developing countries. 7

11 increases the capacity of a country leader to shape foreign policy, within a given year. 6 Assuming that foreign policy is high politics and primarily influenced by the person running the executive branch, we define the head of the executive as the country leader. In a next step, we use information on leadership changes in recipient and donor countries to construct dyadic leader changes. Our units of analysis are leader dyads. To illustrate this approach, consider that President Barack Obama and President Dilma Rousseff had formed the dyad between the United States and Brazil until May 12, We would code a dyadic leadership change resulting in a new dyad between Brazil and the United States for 2016 if her successor, Michel Temer, lasts in office for almost the remaining part of the year, or if Barack Obama surprisingly would have resigned before June 30, Our analysis includes 133 recipient countries that in tandem with the G7 donor countries 7 form 686 country dyads that engage in development cooperation over the period. The panel is unbalanced since some recipient countries enter after Similarly, some donors only engage in development cooperation with a selected set of recipients. Given these limitations, our dataset includes 7505 donor-recipient-leader-pairs and 5010 dyadic leader changes. The average leader dyad lasts about six years. By construction, the shortest period is one year. The most durable leader dyads are between Germany under Chancellor Helmut Kohl and several recipient countries with a duration of 16 years; the exact time Kohl was in office. All G7 countries form administration dyads lasting longer than 10 years, with the exception of the United States, due to presidential term limits. The second independent variable is the change in foreign policy alignment between countries. To proxy the intensity of bilateral relations, we use voting alignment in the United Nations General Assembly. Focusing on the UNGA has several advantages: data availability is generally very high because all sovereign countries have voting rights. Votes in the UNGA furthermore cover a wide array of issues that allow to proxy general alignment tendencies instead of ad hoc political liaisons (Mattes, Leeds, and Carroll, 2015). Voting alignment has thus often been used to proxy political closeness. We measure voting alignment changes as the difference in the percentage of common yes and no votes between any two countries in one administration dyad vote in line with each other between t 1 and t (Thacker, 1999; Faye and Niehaus, 2012). The data is provided by Voeten (2013). Although this difference ranges empirically from -94% to +67%, such radical changes in bilateral relations are rather uncommon. Nevertheless, we test whether our results are sensitive to radical changes by restricting the scope of the alignment change in the robustness section. In addition, we make use of different measures that also include vote abstentions (Barro and Lee, 2005). Note also that Häge and Hug (2016) show that UNGA affinity scores are sensitive to the inclusion of consensus votes that sys- 6 This approach differs from that of Mattes, Leeds, and Carroll (2015) who always use information on the leader who is in power in December for the whole year. 7 Canada, France, Germany, Great Britain, Italy, Japan, and the United States. 8

12 tematically increase voting alignment between all country pairs. As we use changes in voting alignment, this should however not affect our measure if the number of consensus votes does not change dramatically. In the main models, we use all votes since general foreign policy preferences are arguably more reliably revealed by all votes, as compared to only important votes (Andersen, Hansen, and Markussen, 2006). Nonetheless, we test the robustness of our results and also include regular votes votes that reoccur over UNGA sessions and key votes (Kilby, 2009; Kersting and Kilby, 2016). To isolate initial changes in foreign policy alignment from general long- and short-term alignment or dis-alignment tendencies between donor and recipient over time, we further include two variables into our baseline specification: in line with Faye and Niehaus (2012), we control for alignment between the former recipient and donor leader. For this we use average alignment over the past administration dyad instead of recipient leader dyads. This limits the maximum average alignment to 16 years, whereas Faye and Niehaus (2012) have cases where the alignment is averaged over nearly the entire sample period. For instance, Muammar al-gaddafi ruled Libya from 1977 to 2011 and essentially covered the whole spectrum of political relationships with several G7 countries over those years. We argue that our dyadic measure of past alignment is better able to capture past alignment because it does not blur the current relations by relations from decades ago that, in addition, were established by other administrations in donor countries. The effect of past mean alignment thus captures how well the previous administration dyad has worked with each other in general and explains path dependency in current bilateral relations. Moreover, we also include the lagged alignment level since it mechanically determines the possible range of re-alignment. This also controls for short-term alignment effects which could be due to the voting agenda (Häge and Hug, 2016). 4 Methods and Findings In our baseline specification (see equation 1) we regress the natural logarithm of ODA commitments at time t between leader pair d of donor country j to recipient country i on dyadic leader change, alignment changes and their interaction. The alignment change is defined as the difference in common votes between two countries from t 1 and t. The coefficient of interest is the interaction between leader change and changes in voting alignment, i.e. the corresponding change in voting alignment in the UNGA from t 1 (the year of the last leader in either one of the two countries) to t (the first year of the new leader in either one of the two countries). We expect a positive interaction effect of θ implying that alignment following a change in leadership increases aid flows, while dis-alignment decreases aid flows. φ captures the effect of the lagged alignment. As such, it controls for the contemporaneous level of UNGA alignment in a dyad d, which determines the possible range of the change in voting alignment. ψ controls for the past mean alignment of 9

13 the previous administration dyad, to capture the overall relations between the two countries. α ij are donor-recipient fixed effects capturing unobserved time-invariant heterogeneity for specific country dyads. Additionally, γ t are year fixed effects to control for any global shocks that simultaneously affect alignment, leader change and aid commitments across all countries. (1) lnoda ijdt β leader ijdt ` δ alignment ijdt ` θ leader ijdt alignment ijdt ` φ alignment ijt 1 ` ψ meanalignment ijd 1 ` α ij ` γ t ` ɛ ijt Table 1 displays the results of this empirical strategy, when phasing in the different components of the regression model. Column 1 only includes dyadic leadership change. It shows that there is no unconditional effect of leadership turnover on ODA commitments from donor to recipient in a given donor-recipient pair; β is not statistically significant. Hence, the pooled leader change effect from either recipient or donor country does not affect aid allocation in a systematic way. In column 2, we only include the yearly change of voting alignment in the UNGA. The statistically insignificant effect here highlights that yearly fluctuations in political proximity have no effect on their own. This result is in line with previous studies that have argued that yearly alignment changes do not change the general relationship between countries (e.g., Faye and Niehaus, 2012). In column 3, we include our main independent variable the interaction between changes in political alignment and leadership change. Dyadic leader changes with constant bilateral relations as well as yearly fluctuations in alignment in years without leadership turnover are both statistically insignificant. To the contrary, the interaction term is, as expected, positive and statistically significant. Voting convergence after either a donor or recipient leader change is rewarded with more ODA commitments, while divergence is punished with aid cutbacks. Note that the effect of the last year s alignment vanishes as soon as we introduce the interaction term, highlighting that the alignment of the previous leader does not affect ODA commitments that new leaders attract. These findings show that leadership turnover itself does not change aid allocation patterns. Change in leadership only becomes consequential if it changes the trajectory of foreign relations between countries. In column 4, we further exploit the dyadic structure of our data by employing donor-recipient-pair, donor-year and recipient-year fixed effects. This gravity-like approach (Head and Mayer, 2014) enables us to control for other factors that vary on either donor or recipient countries over time and explain ODA allocation. Hence, unobserved heterogeneity is reduced to variables that vary within the dyads over time and are not explained by variables varying over donor and recipient by year, such as GDP or population size. A further benefit of this approach is that we do not decrease our sample size due to data availability. 8 The results show that the magnitude of the θ even increases when controlling for donor and recipient-specific 8 A downside of this approach is that we cannot draw conclusions regarding the effect of leader change in instances where voting alignment is constant, since the fixed effects absorb this variable. 10

14 Table 1: Dyadic Leader Changes Dependent variable: ln ODA commitments (1) (2) (3) (4) Dyadic Leader change (0.024) (0.023) Alignment change (0.187) (0.204) (0.350) Leader change * realignment 0.938*** 1.287*** (0.296) (0.382) Last year alignment 0.412** 0.589** * (0.197) (0.293) (0.294) (0.528) Past mean alignment 0.550** ** 0.832* (0.274) (0.286) (0.309) (0.446) # of observations # of dyads Adjusted R-squared Donor-recipient FE Yes Yes Yes Yes Year FE Yes Yes Yes No Donor-year FE No No No Yes Recipient-year FE No No No Yes Administration change variable in column 5 omitted due to fixed effects. Robust standard errors in parentheses, clustered on donor-recipient dyad. Significance level: *** pă0.01, ** pă0.05, * pă0.1. factors. Hence, we are highly confident that our initial results are not driven by omitted variable bias and if they are, they represent a lower bound estimate. In a next step, we investigate the conditional effect of leadership change and foreign policy realignment on the allocation of ODA commitments by differentiating between foreign policy changes that emanate either after a recipient or donor leader change (see equation 2). 9 The results are displayed in table 2. lnoda ijdt β 1 recipient idt ` β 2 donor jdt ` δ alignment ijdt (2) ` θ 1 recipient jdt alignment ijdt ` θ 2 donor jdt alignment ijdt ` φ alignment ijt 1 ` ψ meanalignment ijd 1 ` α ij ` γ t ` ɛ ijt Column 1 illustrates that the specific type of leader change matters for aid allocation. While changes in donor countries are statistically insignificant, leadership turnover in recipient countries leads to less ODA on average. Taken at face value, this would imply that donors punish heads of executive that take over power in recipient countries. However, the results in column 3 qualify this effect. The interactions between voting alignment change and both recipient and donor leader change are positive and statistically significant. Furthermore, the sizable interaction effect offsets the negative effect of recipient leader change with zero voting alignment changes. Hence, convergence gets rewarded while divergence leads to a reduction in ODA commitments, regardless whether voting re-alignment is a reaction of recipi- 9 Note that β and θ have been changed to β 1 and β 2 as well as θ 1 and θ 2. Although theoretically possible, we do not include mutual leader changes as a separate category because they are empirically rather infrequent. 11

15 Table 2: Dis-aggregate Leader Changes Dependent variable: ln ODA commitments (1) (2) (3) (4) Recipient change *** *** (0.031) (0.031) Donor change (0.026) (0.026) Alignment change (0.187) (0.206) (0.350) Recipient change * realignment 1.198*** 1.185** (0.386) (0.502) Donor change * realignment 0.642** 0.876* (0.315) (0.472) Last year alignment 0.402** 0.589** * (0.197) (0.293) (0.295) (0.529) Past mean alignment 0.548** ** 0.723* (0.273) (0.286) (0.301) (0.434) # of observations # of dyads Adjusted R-squared Donor-recipient FE Yes Yes Yes Yes Year FE Yes Yes Yes No Donor-year FE No No No Yes Recipient-year FE No No No Yes Leader change variables in column 5 omitted due to fixed effects. Robust standard errors in parentheses, clustered on donor-recipient dyad. Significance level: *** pă0.01, ** pă0.05, * pă0.1. ent countries to a new leader in a donor country or a re-alignment of foreign policy after domestic leader change. 10 Furthermore, the effects hold even in the most conservative estimation approach, in which we include donor-recipient-pair, donor-year and recipient-year fixed effects (column 4). 11 Taken together, these results strongly support hypothesis 1 and 2. How consequential are these effects for recipient s revenue streams? To answer this question, we estimate the predicted change of ODA commitments in percentage points with respect to the change in voting alignment effect and the type of leadership turnover (based on model 3 in table 2). The results are plotted in figure 1. At the mean alignment change, representing marginal dis-alignment (see table A1), new recipient leaders receive 9.7% less ODA commitments in their first year. In the opposite case of donor leader change, they receive 3.8% higher ODA commitments. 10 To test for autocorrelation, we reran the models in table 3 including lagged ODA commitments (results not reported). The lagged commitments are statistically significant, and have a point estimate of A test for first order autocorrelation (Wooldridge, 2010; Drukker, 2003) cannot reject the null of no autocorrelation. Furthermore, a Fisher-test for a unit root in panel data using the Dickey-Fuller approach (Choi, 2001), utilizing up to 3 lags, also neglects the presence of a unit root. We also included donor and recipient change and their respective interactions in separate regressions (results not reported). This leads to an increase in the magnitude and statistical significance of the single effects. Hence, our results are not driven by the inclusion of both types of changes. 11 The results are also robust to different forms of clustering (Cameron, Gelbach, and Miller, 2011), such as clustering on donor and recipient or pair and year. 12

16 Figure 1: Marginal Effect of Alignment Change after Leader Change If a newly inaugurated recipient leader, however, chooses to dis-align by one standard deviation which is approximately an 8 percentage point decrease in voting alignment from one year to another ODA commitments to this country shrink by 19.6%. Hence, decreasing political proximity with donor countries in the UNGA increases the negative effect of domestic leader change by about 10 percentage points for aid recipients. In case of donor leader change, dis-alignment seems to have no substantial effect. Conversely, foreign policy convergence gets rewarded with additional aid. A move towards the donor by one standard deviation results in 9.1% more ODA commitments. In substantial terms, these numbers show that signaling political closeness or animosity matters a great deal in times of high uncertainty in bilateral relations, especially with regard to the economic implications of politically granted development aid. Consider for example that the median aid recipient in our sample receives around $100m in development aid from the G7 annually. According to our results, if a new recipient country leader were to alter their foreign policy proximity to international aid providers by one standard deviation, the country would face a cut of 19.6%, i.e. almost $20m. Note further that the size of the alignment change effect is much more pronounced in case of recipient leader change than for donor leader change (see again figure 1). This is due to the fact, that all donor countries react to recipient leader change at once, while only the affected donor reacts after donor leader change. This is also evident from figure 1. The slope of the recipient interaction (black line) is 13

17 much steeper than the slope of the donor interaction (grey line). Nevertheless, donor changes become more important for recipients the more they depend on the donor s aid commitments. Summing up, political re-alignment after leader change is highly consequential for recipient countries. While new recipient leaders can mainly forgo cutbacks by aligning themselves with donors, existing recipient country leaders have an opportunity to fill the public purse when a new donor leader enters office. How lasting is the conditional alignment effect? To explore the time structure, we re-estimate our baseline model (equation 2) using several leads and lags (see table 3). 12 We expect that past leader changes, interacted with the alignment change, do not affect ODA commitments. In contrast, we expect that the effect of conditional voting alignment change should phase out over time, as countries in the same leader dyad grow increasingly familiar with each other. effect of the first impression should lose its power over time. Ergo, the uncertainty-reducing Columns 1 and 2 of table 3 show that there is no effect of prospective recipient leader change on ODA provision. Yet, there is an effect of donor leader change on ODA commitments one and two years before the change occurs. The interaction terms are insignificant in all lead specifications. Thus, the conditional alignment effect cannot be explained by developments before any leader change. Column 3 shows the previous results from the contemporaneous effect of our interactions on ODA commitments. Columns 4 and 5 report the results regarding the ODA provision for the first two years after the leader change. Most importantly, the interaction term between recipient change and alignment change is still positive and statistically significant. Given that the size of the effect decreases one year and vanishes two years after the leader change, we conclude that the effectiveness of conditional voting alignment change decreases over time. The interaction effect for donor change is statistically significant at t, yet fades away within one year. The fact that the effect of last year s alignment is only positive and statistically significant in the year after leadership turnover further underpins the importance of first impressions that subsequently predetermine aid allocation trajectories. Finally, we investigate how different institutional settings and types of leader transitions affect the alignment mechanism. We start by differentiating between legal and illegal leadership change, where we code the latter as irregular entries into office, for example via coups (Goemans, Gleditsch, and Chiozza, 2009). We do so only for recipient countries, as there are no illegal changes in the G7 countries. The results in column 1 of table 4 show a positive and statistically significant alignment change effect in both cases. Furthermore a t-test does not reject that the coefficients are equal. In column 2, we interact our model with a proxy for political struggle, operationalized as years during which a country has more than three heads of executives. In such cases, the alignment change interaction becomes insignificant. This might point to the fact that donors are incapable of gaining enough information dur- 12 Because the median duration of the leader pair dyad is five years, we use two years prior to and after each leader change, in addition to the contemporaneous specification. 14

18 Table 3: Timing of the Conditional Alignment Effect Dependent variable: ln ODA commitments 2 years 1 year leader 1 year 2 years prior prior change after after Recipient change *** *** (0.030) (0.031) (0.031) (0.033) (0.032) Donor change 0.208*** 0.077*** (0.027) (0.029) (0.026) (0.028) (0.026) Alignment change ** 0.437** (0.216) (0.200) (0.206) (0.209) (0.221) Recipient change * realignment *** 0.958** (0.382) (0.415) (0.386) (0.427) (0.496) Donor change * realignment ** (0.378) (0.332) (0.315) (0.360) (0.350) Last year alignment ** 0.992*** (0.302) (0.283) (0.295) (0.325) (0.351) Past mean alignment 0.808** 0.981*** 0.637** * (0.331) (0.309) (0.301) (0.291) (0.298) # of observations # of dyads Adjusted R-squared Donor-recipient FE Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Robust standard errors in parentheses, clustered on donor-recipient dyad. Significance level: *** pă0.01, ** pă0.05, * pă0.1. ing very short executive tenures in recipient states. Thus, they are unable to figure out who they are dealing with and thus revert to some standard aid allocation. In column 3, we test whether domestic-support-group change in addition to leader-change amplifies the effects from changes in voting alignment. Mattes, Leeds, and Carroll (2015) highlight that changes in the domestic support groups are the main driver of significant foreign policy re-alignment. We adapt their specification by including their core set of control variables in our dyadic setting (see table A2 in the appendix). We find evidence in favor of our argument regardless of a simultaneous change in the support group of the leader the interaction term is positive and statistically significant in both cases. At first glance the magnitude of the point estimate is higher in the case of domestic support group change. The t-test, however, indicates no difference between the coefficients. Thus, alterations in the conditions surrounding leader change seem not to change the bigger picture. settings. We further differentiate between different institutional as well as time specific In column 4, we adapt the Dreher and Jensen (2013) specification and differentiate between the Cold War and post-cold War period, but use all votes in the UNGA instead of focusing on key votes alone. In column 5, we subdivide the sample into democracies and autocracies in line with Carter and Stone (2015). 13 The interaction terms between leader change and changes in political proximity show 13 Due to space restrictions, we do not report the coefficients of the additional control variables (see table A2 in the appendix). They are, however, in line with the findings of previous research. 15